Ziyu Wang

Ziyu Wang, Yucen Luo, Yueru Li et al.

Many problems in causal inference and economics can be formulated in the framework of conditional moment models, which characterize the target function through a collection of conditional moment restrictions. For nonparametric conditional moment models, efficient estimation often relies on preimposed conditions on various measures of ill-posedness of the hypothesis space, which are hard to validate when flexible models are used. In this work, we address this issue by proposing a procedure that automatically learns representations with controlled measures of ill-posedness. Our method approximates a linear representation defined by the spectral decomposition of a conditional expectation operator, which can be used for kernelized estimators and is known to facilitate minimax optimal estimation in certain settings. We show this representation can be efficiently estimated from data, and establish L2 consistency for the resulting estimator. We evaluate the proposed method on proximal causal inference tasks, exhibiting promising performance on high-dimensional, semi-synthetic data.

Wenkai Tan, Justus Renkhoff, Alvaro Velasquez et al.

Deep Learning (DL) and Deep Neural Networks (DNNs) are widely used in various domains. However, adversarial attacks can easily mislead a neural network and lead to wrong decisions. Defense mechanisms are highly preferred in safety-critical applications. In this paper, firstly, we use the gradient class activation map (GradCAM) to analyze the behavior deviation of the VGG-16 network when its inputs are mixed with adversarial perturbation or Gaussian noise. In particular, our method can locate vulnerable layers that are sensitive to adversarial perturbation and Gaussian noise. We also show that the behavior deviation of vulnerable layers can be used to detect adversarial examples. Secondly, we propose a novel NoiseCAM algorithm that integrates information from globally and pixel-level weighted class activation maps. Our algorithm is susceptible to adversarial perturbations and will not respond to Gaussian random noise mixed in the inputs. Third, we compare detecting adversarial examples using both behavior deviation and NoiseCAM, and we show that NoiseCAM outperforms behavior deviation modeling in its overall performance. Our work could provide a useful tool to defend against certain adversarial attacks on deep neural networks.

Ziyu Wang, Wenhao Jiang, Zixuan Zhang et al.

Sequential processes in real-world often carry a combination of simple subsystems that interact with each other in certain forms. Learning such a modular structure can often improve the robustness against environmental changes. In this paper, we propose recurrent independent Grid LSTM (RigLSTM), composed of a group of independent LSTM cells that cooperate with each other, for exploiting the underlying modular structure of the target task. Our model adopts cell selection, input feature selection, hidden state selection, and soft state updating to achieve a better generalization ability on the basis of the recent Grid LSTM for the tasks where some factors differ between training and evaluation. Specifically, at each time step, only a fraction of cells are activated, and the activated cells select relevant inputs and cells to communicate with. At the end of one time step, the hidden states of the activated cells are updated by considering the relevance between the inputs and the hidden states from the last and current time steps. Extensive experiments on diversified sequential modeling tasks are conducted to show the superior generalization ability when there exist changes in the testing environment. Source code is available at \url{https://github.com/ziyuwwang/rig-lstm}.

Yuxuan Wu, Ziyu Wang, Bhiksha Raj et al.

We contribute an unsupervised method that effectively learns disentangled content and style representations from sequences of observations. Unlike most disentanglement algorithms that rely on domain-specific labels or knowledge, our method is based on the insight of domain-general statistical differences between content and style -- content varies more among different fragments within a sample but maintains an invariant vocabulary across data samples, whereas style remains relatively invariant within a sample but exhibits more significant variation across different samples. We integrate such inductive bias into an encoder-decoder architecture and name our method after V3 (variance-versus-invariance). Experimental results show that V3 generalizes across multiple domains and modalities, successfully learning disentangled content and style representations, such as pitch and timbre from music audio, digit and color from images of hand-written digits, and action and character appearance from simple animations. V3 demonstrates strong disentanglement performance compared to existing unsupervised methods, along with superior out-of-distribution generalization under few-shot adaptation compared to supervised counterparts. Lastly, symbolic-level interpretability emerges in the learned content codebook, forging a near one-to-one alignment between machine representation and human knowledge.

Ziyu Wang, Yuhao Zhou, Jun Zhu

We investigate nonlinear instrumental variable (IV) regression given high-dimensional instruments. We propose a simple algorithm which combines kernelized IV methods and an arbitrary, adaptive regression algorithm, accessed as a black box. Our algorithm enjoys faster-rate convergence and adapts to the dimensionality of informative latent features, while avoiding an expensive minimax optimization procedure, which has been necessary to establish similar guarantees. It further brings the benefit of flexible machine learning models to quasi-Bayesian uncertainty quantification, likelihood-based model selection, and model averaging. Simulation studies demonstrate the competitive performance of our method.

Yue Xu, Yong-Lu Li, Kaitong Cui et al.

Data-efficient learning has garnered significant attention, especially given the current trend of large multi-modal models. Recently, dataset distillation has become an effective approach by synthesizing data samples that are essential for network training. However, it remains to be explored which samples are essential for the dataset distillation process itself. In this work, we study the data efficiency and selection for the dataset distillation task. By re-formulating the dynamics of distillation, we provide insight into the inherent redundancy in the real dataset, both theoretically and empirically. We propose to use the empirical loss value as a static data pruning criterion. To further compensate for the variation of the data value in training, we find the most contributing samples based on their causal effects on the distillation. The proposed selection strategy can efficiently exploit the training dataset, outperform the previous SOTA distillation algorithms, and consistently enhance the distillation algorithms, even on much larger-scale and more heterogeneous datasets, e.g., full ImageNet-1K and Kinetics-400. We believe this paradigm will open up new avenues in the dynamics of distillation and pave the way for efficient dataset distillation. Our code is available on https://github.com/silicx/GoldFromOres-BiLP.

Ziyu Wang, Yanjie Ze, Yifei Sun et al.

Learning policies that can generalize to unseen environments is a fundamental challenge in visual reinforcement learning (RL). While most current methods focus on acquiring robust visual representations through auxiliary supervision, pre-training, or data augmentation, the potential of modern vision foundation models remains underleveraged. In this work, we introduce Segment Anything Model for Generalizable visual RL (SAM-G), a novel framework that leverages the promptable segmentation ability of Segment Anything Model (SAM) to enhance the generalization capabilities of visual RL agents. We utilize image features from DINOv2 and SAM to find correspondence as point prompts to SAM, and then SAM produces high-quality masked images for agents directly. Evaluated across 8 DMControl tasks and 3 Adroit tasks, SAM-G significantly improves the visual generalization ability without altering the RL agents' architecture but merely their observations. Notably, SAM-G achieves 44% and 29% relative improvements on the challenging video hard setting on DMControl and Adroit respectively, compared to state-of-the-art methods. Video and code: https://yanjieze.com/SAM-G/

Tianyi Zhang, Li Zhang, Zhaoyi Hou et al.

Planning in a text-based environment continues to be a major challenge for AI systems. Recent approaches have used language models to predict a planning domain definition (e.g., PDDL) but have only been evaluated in closed-domain simulated environments. To address this, we present Proc2PDDL , the first dataset containing open-domain procedural texts paired with expert-annotated PDDL representations. Using this dataset, we evaluate state-of-the-art models on defining the preconditions and effects of actions. We show that Proc2PDDL is highly challenging, with GPT-3.5's success rate close to 0% and GPT-4's around 35%. Our analysis shows both syntactic and semantic errors, indicating LMs' deficiency in both generating domain-specific prgorams and reasoning about events. We hope this analysis and dataset helps future progress towards integrating the best of LMs and formal planning.

Gaole Dai, Wei Wu, Ziyu Wang et al.

Learned optimizers are a crucial component of meta-learning. Recent advancements in scalable learned optimizers have demonstrated their superior performance over hand-designed optimizers in various tasks. However, certain characteristics of these models, such as an unstable learning curve, limited ability to handle unseen tasks and network architectures, difficult-to-control behaviours, and poor performance in fine-tuning tasks impede their widespread adoption. To tackle the issue of generalization in scalable learned optimizers, we propose a hybrid-update-based (HUB) optimization strategy inspired by recent advancements in hard prompt tuning and result selection techniques used in large language and vision models. This approach can be easily applied to any task that involves hand-designed or learned optimizer. By incorporating hand-designed optimizers as the second component in our hybrid approach, we are able to retain the benefits of learned optimizers while stabilizing the training process and, more importantly, improving testing performance. We validate our design through a total of 17 tasks, consisting of thirteen training from scratch and four fine-tuning settings. These tasks vary in model sizes, architectures, or dataset sizes, and the competing optimizers are hyperparameter-tuned. We outperform all competitors in 94% of the tasks with better testing performance. Furthermore, we conduct a theoretical analysis to examine the potential impact of our hybrid strategy on the behaviours and inherited traits of learned optimizers.

Ziyu Wang, Yuhao Zhou, Tongzheng Ren et al.

Recent years have witnessed an upsurge of interest in employing flexible machine learning models for instrumental variable (IV) regression, but the development of uncertainty quantification methodology is still lacking. In this work we present a novel quasi-Bayesian procedure for IV regression, building upon the recently developed kernelized IV models and the dual/minimax formulation of IV regression. We analyze the frequentist behavior of the proposed method, by establishing minimax optimal contraction rates in $L_2$ and Sobolev norms, and discussing the frequentist validity of credible balls. We further derive a scalable inference algorithm which can be extended to work with wide neural network models. Empirical evaluation shows that our method produces informative uncertainty estimates on complex high-dimensional problems.

Caglar Gulcehre, Sergio Gómez Colmenarejo, Ziyu Wang et al.

Offline reinforcement learning restricts the learning process to rely only on logged-data without access to an environment. While this enables real-world applications, it also poses unique challenges. One important challenge is dealing with errors caused by the overestimation of values for state-action pairs not well-covered by the training data. Due to bootstrapping, these errors get amplified during training and can lead to divergence, thereby crippling learning. To overcome this challenge, we introduce Regularized Behavior Value Estimation (R-BVE). Unlike most approaches, which use policy improvement during training, R-BVE estimates the value of the behavior policy during training and only performs policy improvement at deployment time. Further, R-BVE uses a ranking regularisation term that favours actions in the dataset that lead to successful outcomes. We provide ample empirical evidence of R-BVE's effectiveness, including state-of-the-art performance on the RL Unplugged ATARI dataset. We also test R-BVE on new datasets, from bsuite and a challenging DeepMind Lab task, and show that R-BVE outperforms other state-of-the-art discrete control offline RL methods.

Konrad Zolna, Alexander Novikov, Ksenia Konyushkova et al.

Behavior cloning (BC) is often practical for robot learning because it allows a policy to be trained offline without rewards, by supervised learning on expert demonstrations. However, BC does not effectively leverage what we will refer to as unlabeled experience: data of mixed and unknown quality without reward annotations. This unlabeled data can be generated by a variety of sources such as human teleoperation, scripted policies and other agents on the same robot. Towards data-driven offline robot learning that can use this unlabeled experience, we introduce Offline Reinforced Imitation Learning (ORIL). ORIL first learns a reward function by contrasting observations from demonstrator and unlabeled trajectories, then annotates all data with the learned reward, and finally trains an agent via offline reinforcement learning. Across a diverse set of continuous control and simulated robotic manipulation tasks, we show that ORIL consistently outperforms comparable BC agents by effectively leveraging unlabeled experience.

Tom Le Paine, Cosmin Paduraru, Andrea Michi et al.

Offline reinforcement learning (RL purely from logged data) is an important avenue for deploying RL techniques in real-world scenarios. However, existing hyperparameter selection methods for offline RL break the offline assumption by evaluating policies corresponding to each hyperparameter setting in the environment. This online execution is often infeasible and hence undermines the main aim of offline RL. Therefore, in this work, we focus on \textit{offline hyperparameter selection}, i.e. methods for choosing the best policy from a set of many policies trained using different hyperparameters, given only logged data. Through large-scale empirical evaluation we show that: 1) offline RL algorithms are not robust to hyperparameter choices, 2) factors such as the offline RL algorithm and method for estimating Q values can have a big impact on hyperparameter selection, and 3) when we control those factors carefully, we can reliably rank policies across hyperparameter choices, and therefore choose policies which are close to the best policy in the set. Overall, our results present an optimistic view that offline hyperparameter selection is within reach, even in challenging tasks with pixel observations, high dimensional action spaces, and long horizon.

Ziyu Wang, Alexander Novikov, Konrad Zolna et al.

Offline reinforcement learning (RL), also known as batch RL, offers the prospect of policy optimization from large pre-recorded datasets without online environment interaction. It addresses challenges with regard to the cost of data collection and safety, both of which are particularly pertinent to real-world applications of RL. Unfortunately, most off-policy algorithms perform poorly when learning from a fixed dataset. In this paper, we propose a novel offline RL algorithm to learn policies from data using a form of critic-regularized regression (CRR). We find that CRR performs surprisingly well and scales to tasks with high-dimensional state and action spaces -- outperforming several state-of-the-art offline RL algorithms by a significant margin on a wide range of benchmark tasks.

Bin Dai, Ziyu Wang, David Wipf

In narrow asymptotic settings Gaussian VAE models of continuous data have been shown to possess global optima aligned with ground-truth distributions. Even so, it is well known that poor solutions whereby the latent posterior collapses to an uninformative prior are sometimes obtained in practice. However, contrary to conventional wisdom that largely assigns blame for this phenomena on the undue influence of KL-divergence regularization, we will argue that posterior collapse is, at least in part, a direct consequence of bad local minima inherent to the loss surface of deep autoencoder networks. In particular, we prove that even small nonlinear perturbations of affine VAE decoder models can produce such minima, and in deeper models, analogous minima can force the VAE to behave like an aggressive truncation operator, provably discarding information along all latent dimensions in certain circumstances. Regardless, the underlying message here is not meant to undercut valuable existing explanations of posterior collapse, but rather, to refine the discussion and elucidate alternative risk factors that may have been previously underappreciated.

Konrad Zolna, Scott Reed, Alexander Novikov et al.

We show that a critical vulnerability in adversarial imitation is the tendency of discriminator networks to learn spurious associations between visual features and expert labels. When the discriminator focuses on task-irrelevant features, it does not provide an informative reward signal, leading to poor task performance. We analyze this problem in detail and propose a solution that outperforms standard Generative Adversarial Imitation Learning (GAIL). Our proposed method, Task-Relevant Adversarial Imitation Learning (TRAIL), uses constrained discriminator optimization to learn informative rewards. In comprehensive experiments, we show that TRAIL can solve challenging robotic manipulation tasks from pixels by imitating human operators without access to any task rewards, and clearly outperforms comparable baseline imitation agents, including those trained via behaviour cloning and conventional GAIL.

Tom Le Paine, Caglar Gulcehre, Bobak Shahriari et al.

This paper introduces R2D3, an agent that makes efficient use of demonstrations to solve hard exploration problems in partially observable environments with highly variable initial conditions. We also introduce a suite of eight tasks that combine these three properties, and show that R2D3 can solve several of the tasks where other state of the art methods (both with and without demonstrations) fail to see even a single successful trajectory after tens of billions of steps of exploration.

Yuke Zhu, Ziyu Wang, Josh Merel et al.

We propose a model-free deep reinforcement learning method that leverages a small amount of demonstration data to assist a reinforcement learning agent. We apply this approach to robotic manipulation tasks and train end-to-end visuomotor policies that map directly from RGB camera inputs to joint velocities. We demonstrate that our approach can solve a wide variety of visuomotor tasks, for which engineering a scripted controller would be laborious. In experiments, our reinforcement and imitation agent achieves significantly better performances than agents trained with reinforcement learning or imitation learning alone. We also illustrate that these policies, trained with large visual and dynamics variations, can achieve preliminary successes in zero-shot sim2real transfer. A brief visual description of this work can be viewed in https://youtu.be/EDl8SQUNjj0

Serkan Cabi, Sergio Gómez Colmenarejo, Matthew W. Hoffman et al.

This paper introduces the Intentional Unintentional (IU) agent. This agent endows the deep deterministic policy gradients (DDPG) agent for continuous control with the ability to solve several tasks simultaneously. Learning to solve many tasks simultaneously has been a long-standing, core goal of artificial intelligence, inspired by infant development and motivated by the desire to build flexible robot manipulators capable of many diverse behaviours. We show that the IU agent not only learns to solve many tasks simultaneously but it also learns faster than agents that target a single task at-a-time. In some cases, where the single task DDPG method completely fails, the IU agent successfully solves the task. To demonstrate this, we build a playroom environment using the MuJoCo physics engine, and introduce a grounded formal language to automatically generate tasks.

Ziyu Wang, Josh Merel, Scott Reed et al.

Deep generative models have recently shown great promise in imitation learning for motor control. Given enough data, even supervised approaches can do one-shot imitation learning; however, they are vulnerable to cascading failures when the agent trajectory diverges from the demonstrations. Compared to purely supervised methods, Generative Adversarial Imitation Learning (GAIL) can learn more robust controllers from fewer demonstrations, but is inherently mode-seeking and more difficult to train. In this paper, we show how to combine the favourable aspects of these two approaches. The base of our model is a new type of variational autoencoder on demonstration trajectories that learns semantic policy embeddings. We show that these embeddings can be learned on a 9 DoF Jaco robot arm in reaching tasks, and then smoothly interpolated with a resulting smooth interpolation of reaching behavior. Leveraging these policy representations, we develop a new version of GAIL that (1) is much more robust than the purely-supervised controller, especially with few demonstrations, and (2) avoids mode collapse, capturing many diverse behaviors when GAIL on its own does not. We demonstrate our approach on learning diverse gaits from demonstration on a 2D biped and a 62 DoF 3D humanoid in the MuJoCo physics environment.

Nicolas Heess, Dhruva TB, Srinivasan Sriram et al.

The reinforcement learning paradigm allows, in principle, for complex behaviours to be learned directly from simple reward signals. In practice, however, it is common to carefully hand-design the reward function to encourage a particular solution, or to derive it from demonstration data. In this paper explore how a rich environment can help to promote the learning of complex behavior. Specifically, we train agents in diverse environmental contexts, and find that this encourages the emergence of robust behaviours that perform well across a suite of tasks. We demonstrate this principle for locomotion -- behaviours that are known for their sensitivity to the choice of reward. We train several simulated bodies on a diverse set of challenging terrains and obstacles, using a simple reward function based on forward progress. Using a novel scalable variant of policy gradient reinforcement learning, our agents learn to run, jump, crouch and turn as required by the environment without explicit reward-based guidance. A visual depiction of highlights of the learned behavior can be viewed following https://youtu.be/hx_bgoTF7bs .

Josh Merel, Yuval Tassa, Dhruva TB et al.

Rapid progress in deep reinforcement learning has made it increasingly feasible to train controllers for high-dimensional humanoid bodies. However, methods that use pure reinforcement learning with simple reward functions tend to produce non-humanlike and overly stereotyped movement behaviors. In this work, we extend generative adversarial imitation learning to enable training of generic neural network policies to produce humanlike movement patterns from limited demonstrations consisting only of partially observed state features, without access to actions, even when the demonstrations come from a body with different and unknown physical parameters. We leverage this approach to build sub-skill policies from motion capture data and show that they can be reused to solve tasks when controlled by a higher level controller.

Ziyu Wang, Victor Bapst, Nicolas Heess et al.

This paper presents an actor-critic deep reinforcement learning agent with experience replay that is stable, sample efficient, and performs remarkably well on challenging environments, including the discrete 57-game Atari domain and several continuous control problems. To achieve this, the paper introduces several innovations, including truncated importance sampling with bias correction, stochastic dueling network architectures, and a new trust region policy optimization method.

Ziyu Wang, Tom Schaul, Matteo Hessel et al.

In recent years there have been many successes of using deep representations in reinforcement learning. Still, many of these applications use conventional architectures, such as convolutional networks, LSTMs, or auto-encoders. In this paper, we present a new neural network architecture for model-free reinforcement learning. Our dueling network represents two separate estimators: one for the state value function and one for the state-dependent action advantage function. The main benefit of this factoring is to generalize learning across actions without imposing any change to the underlying reinforcement learning algorithm. Our results show that this architecture leads to better policy evaluation in the presence of many similar-valued actions. Moreover, the dueling architecture enables our RL agent to outperform the state-of-the-art on the Atari 2600 domain.

Yishu Miao, Ziyu Wang, Phil Blunsom

This paper presents novel Bayesian optimisation algorithms for minimum error rate training of statistical machine translation systems. We explore two classes of algorithms for efficiently exploring the translation space, with the first based on N-best lists and the second based on a hypergraph representation that compactly represents an exponential number of translation options. Our algorithms exhibit faster convergence and are capable of obtaining lower error rates than the existing translation model specific approaches, all within a generic Bayesian optimisation framework. Further more, we also introduce a random embedding algorithm to scale our approach to sparse high dimensional feature sets.